can't build afni with R about neurodocker HOT 7 CLOSED

Greetings+

Manually editing the recipe from
from
rPkgsInstall -pkgs ALL
to
/opt/afni-latest/rPkgsInstall -pkgs ALL seems to have done the trick!

The error you get might also be related to singularity - it could be that building this with docker would actually work.
Unfortunately my cluster only support Singularity :(

If you just want an AFNI container with all required dependencies and R-packages installed and tested, you can also use the singularity container from the Neurodesk project (https://www.neurodesk.org/docs/getting-started/neurocontainers/singularity/)

Amazing! Thx I see that it contains python too!

from neurodocker.

I think we can close this issue here. As discussed in this issue (NeuroDesk/neurocontainers#505), the problems @andreifoldes experienced come from running the singularity container without --cleanenv and R-packages from the host environment interfered with the container environment.

from neurodocker.

It seems that the R install of AFNI is not something we test regularly: https://github.com/ReproNim/neurodocker/actions/runs/6520678281/job/17708466328

We should probably fix that.

That being said AFNI is the most "broken" install that neurodocker has at the moment so any help to improve that is most welcome

from neurodocker.

Dear @andreifoldes,

You can have a look at this neurodocker recipe, which works for installing the R-packages in AFNI:
https://github.com/NeuroDesk/neurocontainers/blob/master/recipes/afni/build.sh

In particular, you could try fedora 35 as a base image, which I have found to work best for AFNI:

   --base-image fedora:35 \
   --pkg-manager yum \
   --afni version=latest method=binaries install_r_pkgs='true' install_python3='true'

The error you get might also be related to singularity - it could be that building this with docker would actually work. It seems that it can't find the executable rPkgsInstall on the PATH, which could be a path error that somehow works in Docker?

If you just want an AFNI container with all required dependencies and R-packages installed and tested, you can also use the singularity container from the Neurodesk project (https://www.neurodesk.org/docs/getting-started/neurocontainers/singularity/):

curl -X GET https://objectstorage.us-ashburn-1.oraclecloud.com/n/sd63xuke79z3/b/neurodesk/o/afni_23.0.07_20230302.simg -O

I hope that helps
Kind regards
Steffen

from neurodocker.

I stand corrected... while adding full path in the recipe does get the build issue fixed, once actually trying it out on real life data produces:

Loading required package: lme4
Loading required package: Matrix
Error: package or namespace load failed for 'lme4':
 package 'lme4' was installed before R 4.0.0: please re-install it
Failed with error:  'package 'lme4' could not be loaded'

This occurs with the neurodesk simg file specifically!

In terms of neurodocker - can yum install specific version of R? I have a feeling that if only I could get R 3.6 all 3dLMEr would work.

from neurodocker.

Dear @andreifoldes,

neurodocker cannot install a specific version of R through yum - the best way I found to get a specific version of R is to choose a baseimage version that brings this R version. Another option could be to install R manually in a specific version.

So, 3dLMEr only works with R 3.6? Is that correct?

from neurodocker.

choose a baseimage version that brings this R version

I see! Can you recommend which baseimage would be it for R 3.6, how could I check?

So, 3dLMEr only works with R 3.6? Is that correct?

I haven't checked this in all Linux distributions, but we have Centos7 on our cluster and there I had trouble getting lme4 working with r-base>=4 and had better success with 3.6. 3dLMEr needs lme4, lmerTest, phia and snow to work. In terms of singularity image I would be happy to have any working solution ^^

Another option could be to install R manually in a specific version.

Right, but then I would need to install it every time I call the function right? Or would you turn it into a sandbox?

from neurodocker.